Thread 511985976

>saying it could be a one-off invention like the lightbulb
...what?

>>511985976 (OP)
>>511986125

>one-off invention like the lightbulb
wut?

>>511986125
it's basically the ultimate general purpose tool, so of course this is how checked out boomers at the Fed think. Reminds me of Krugman saying in 1998 that the internet was a nothingburger and would be less impactful than the fax machine.

>>511986230
>>511986271
beat me too it. It's unreal how fucking stupid economists are.

lol this reminds me of when the Fed making fun of CBDC (picrel)

>>511986125

wasn't the light bulb transformative and changed society ?

>HAHAHAHAHAHAHAHA ...
lol

>>511985976 (OP)
Actually, automated factories wont need light bulbs.

>>511985976 (OP)
>Out of touch Boomers say stupid shit again about things they don't understand combined with liberal lack of imagination and thought killing nihilism film at 11

>>511986467
Maybe? It was more everything that went along with the lightbulb that was transformative. People had light.

I am also having this thread >>>511982142 on Trump demanding Powell be overtaken by the board of governors

fucking kek
i think i'll have a beer

>>511985976 (OP)
I think it's more than that, but companies are going to fuck up implementing it and humans will need to fix it. It is essentially the next "internet" in that it will power robots and other utilities that serve what humans have used engineering for since the dawn of time: to make life easier. They're spending insane amounts of capex and signing researchers to professional athlete-type contracts kek. There will be a bubble, but it'll just be a fart; AI isn't going anywhere just like the Internet didn't go anywhere in the late 90s; it evolved.

How much did lightbulbs increase worker "productivity"? Lightbulbs made lighting the interiors of buildings enormously easier and more convenient and changed the way we live, but how much more "productive" did that make people? I legitimately don't know.

>>511986792
Exactly, this is boomer senility at its finest.

>>511986719
The orange nigger fucked the market today and is absolutely seething kek.

>>511986467
Yes, it had huge consequences
The night became a far more accessible time that it was with combustion lamps

>>511986318
Ivory tower syndrome, they live in echo chambers, like a physical version of reddit back in the day (Princeton)
DEI and other liberal bullshit didn't spring up overnight.

>>511985976 (OP)
That is the most nigger retarded take I have ever heard. The lightbulb really nigger.

I love it when /pol/ shills are forced to side with the fed and laura loomer, lmao

>>511985976 (OP)
Just a one off invention... like writing...

>>511986125
They're lizard people, they dont need bulbs besides those heated ones while they're indoors

>guys, listen, the economic impact of AI could be as meaningless as electricity
>>511987074
kek

>>511987074
Like language

>>511987009
Who here is siding with the Fed? And what does Loomer have to do with any of this saar?

<<<-----hey everyone, this is about this paper by the Fed on AI
https://www.federalreserve.gov/econres/feds/files/2025053pap.pdf

>>511987174
>saar

this will be interesting

>>511987269
nm, I misunderstood your post

>>511986125
>>511986232
>>511986467
>To return to the light-bulb metaphor, they write that “some inventions, such as the light bulb, temporarily raise productivity growth as adoption spreads, but the effect fades when the market is saturated; that is, the level of output per hour is permanently higher, but the growth rate is not.”
The idea is that AI is really just a tool to calculate or organize things faster which means it will eventually be a market that gets flooded with multiple alternatives, this is why OpenAI almost shat itself over Deepseek. If it reaches that point, and it becomes normalized like having a computer and calculator, it will probably have a rapid growth (where the big players suddenly get rich) and then even off.

They’re skeptical that GENERATIVE AI will take off to the extent that people claim it will, generative AI seems overblown in comparison to having a chatbot on hand. Sloppa is inherently soulless IMO

>>511986792
>Extend the entire human activity cycle by 12 hours out of darkness affordably and conveniently.
>How could this possibly have impacted productivity?

I get that you want to give the fed the benefit of the doubt because they're big government and that makes you feel safe but goddamn son.

>Federal Reserve Board staff paper

I don't know how much they have to do with the Chairman but suddenly a firing sweep seems entirely reasonable.

>>511987365
>suddenly a firing sweep seems entirely reasonable.

>>511985976 (OP)
It might be a more prescient idea than it first appears. Early concepts of gas and electric lighting was mulled around in early 1800s but not widely commercial until 1870

>>511987326
You don't even know what generative AI means.

kek

and also this, too

>>511987626
Asking people to define what generative AI or AGI produces similar hand waving to asking what "woke right" means

>>511986230
He was right. Global growth since around 2000 was a mirage created by inflation. Numbers got bigger, but purchasing power stayed flat or went down. Nominal returns are one hell of a drug.

>>511988150
Mercury dimes status?

>>511988150
Not even by your own standard because in 1998 a vampire hunter D vhs could run 60-80 dollars new in 1998 dollars, now it's a free media attached to most streaming services if you're not pirating it over the internet in the first place. This discounts value produced by gps, zoom meetings, real time weather and power reporting etc.

There is no limit to liberal lack of intelligence or imagination because you weren't including media in your purchasing power and probably still thinking about gold coins and suits or other boomer lessons.

Just die already, please, for the good of us all.

>Here come arguments about food production etc
>Here come my counters that AI is going to streamline every other market

Every one of you commies wants your star trek utopia in which UBI and other near magical ideas will actually work yet hilariously you piss on AI which is a major step towards that. You're not even practically minded for your own sakes.

I will be happy to short your investments and invest opposite of you going forward however.

>>511988150
you're so dense that your asshole turned into a double black hole

the criticisms

>>511988812
Translation: We lack imagination.

Yeah, as I said. The only chokepoint is power requirements and power is an engineering problem that can be solved.
That's even considering making it this public use thing which requires so much power, which is a Meta leaning.
The values of solutions being returned per watt is immense.
An engineering problem that can be solved easier using AI also lol.

>>511988916
They're not wrong.

>>511985976 (OP)
Fucking morons, holy shit.

>>511988916
the rest of the criticism

>>511989037
it's literal damage control because of the AI-unemployment scare
wish we could see the actual report
but doesn't matter anyway, we're gonna see it in RL soon

>>511985976 (OP)
which is correct at this stage. something like 70-80% of all AI projects aren't making a return on investment. it is pure hype. That will change, but currently the fed is correct.

>>511989037
I can tell you don't know shit about "AI", which is just rebranded ML. I've been working with those solutions for years and they have specific use cases, but really replace no one. Example use case, customer had oil wells they sold, they need a way to estimate size of well. ML was able to accurately estimate the size. LLMs are just oversized ML models in a statistical next word machine, which is ultimately nigger technology. They'll start to collapse as artificial data makes it into the training data set and we're already hitting peak return on the tech as is. You're a moron.

>>511986694
>mnost people have nothing to say to each other
if only, one could dream

>>511988916
They're wrong because people are already ditching machine learning searches for GenAI searches in droves.
They're also wrong because CHatGPT is already having an immense effect on productivity, people just aren't admitting it, signing NDAs and such, and companies don't want to admit it either for fear of lawsuits and putting in policies to kid glove it because of copyright implications.

It will amplify as more industries start cross sectoring it. It will not be a slow ramp up like the transistor because it provides solutions to logistical and other problems itself. It's unique in info tech history that way.

These people are drones regurgitating info. AI could replace them easily.
It almost feels like they wanted a case to keep from lowering rates and bullshitted themselves into a worldview to match.

Someone feed the whole paper into ChatGPT now and let AI defend itself lol.

>>511989577
>something like 70-80% of all AI projects aren't making a return on investment
kek, savage non-sequitur
>>511989656
>my dad works at nintendo and

>>511987199
>electric dynamo is an example
well if that's the case so are LED bulbs

>>511989787
>i'm smart but can't refute anything said
Niggers love nigger technology cause it makes them feel smart.

>>511989122
Sure, they might be right about LLMs, but the point is to be the first to achieve a functional AGI, or something close to it. LLMs are only a step on that path.

one of the jobs of the fed is to signal the economy's health, but never too loudly or transparently, so as not to frighten investors.

all they're saying right now is greenspeak for "AI is, factually, a bubble, and investors need to diversify"

>>511989968
AGI isn't coming around in any of our lifetimes nor those of our great grand children.

>>511987326
Remember that time when a lightbulb shined all that light on its foundational principles and made itself, and all other light emitting objects, 2% brighter?

https://deepmind.google/discover/blog/alphaevolve-a-gemini-powered-coding-agent-for-designing-advanced-algorithms/

>>511989968
LLMs are not a step you retard, they're a statistical guess the next word machine. AGI requires actual reasoning, something statistics are not capable of. They have to figure out a completely different way to do things to achieve AGI (real AI). Everything else is ML.

>>511989774
>These people are drones regurgitating info.
lol these are the economists bessent speaks of in pic rel

>>511989948
>you can't refute my completely worthless unrelated invalid made up low-IQ comment
lmao

and kek looks like bessent knows the Fed is broke, too

>>511990083
This is deeply anti-semitic. I'm reporting you to Sam Altman and the ADL.

>>511990028
How would you know?

>>511990083
Yes it is. The goal of AI researchers and companies is to develop an AGI, LLMs are a stepping stone.

>>511989577
>visibility
>transparent background
oh the ironing

>>511990028
>fucking a dog allow canadians to see the future

>>511990411
No but fucking mutts can't see a jewish trick right in front of them to all of our detriment. They aren't murdering people over exposing their lies for no reason. I'm sure that jeet from OpenAI actually killed himself.

>>511990264
Explain specifically how Bayesian statistics magically leads to AGI, sketch it out, mongol rape baby retard.

>>511990563
lmao
>>511990602
>go ahead, write the algorithm for AGI
lmao
/pol/ AI threads are truly amazing...

>>511990168
>The fucking Fed can't manage it's own budget
Yeah I shouldn't be surprised. This should be headline news but we know why it isn't.
Ranking Bessent up there with Tom Homan now, maybe he can adopt more of that take no prisoners stance.

>cue libtards to flood in and defend the fed now lol

>>511990083
Wrongo. You can say the ones we have are bad at it, but all that will ever be needed is a guess the next word machine.

https://en.wikipedia.org/wiki/Solomonoff's_theory_of_inductive_inference

maybe the Fed read this paper that said the longer AI does math the more mistakes it makes

>>511990895
be weary of bessent
bessent wants to lower capital requirements for banks

>>511990896
>wikipedia article with huge problems
Inductive reasoning is the least accurate anyway but you're a dumbass that just googled something quick and pasted it. AGI isn't happening anytime soon, get over it. Your "AI" market is a bubble that will crash.

>>511991097
Considering Trump surrounded himself with Bessent (Soros guy) and Lutnick (Epstein guy), you have to wonder if those are really just his people

>>511990983
Maybe the Fed should keep up with the news then.

https://arxiv.org/abs/2506.13131

>>511990983
>AI will never do hands 2: electric boogaloo
>>511991164
>Your "AI" market is a bubble that will crash
lmao, prove it
short google and nvidia
post the screenshot, big guy

>>511985976 (OP)
Except the lightbulb was massive and untethered society from the day night cycle.

>>511990983
>>511991167
>reeee, this is all wrong, AGI is here tomorrow!!!
It's hilarious watching the average dipshit ChatGPT user trying to pretend they understand shit about how it works then tell everyone what will happen.

in compsci we have a term called techincal debt, it's code in your codebase that was the quick solution at the time, but you pay the costs later because it constrains you and you need to come back later to refactor it
LLM code is pure technical debt. they spit out code that might solve the problem at hand, but you will need to review at some point. the absolute worst case is when it produces output that appears correct but is subtly wrong because a person didn't think it through, a model simply generated the next most likely token.
you can't just give a non-programmer chatgpt and have them vibe code something beyond a to-do notes app effectively. the complexity of the program reaches a tipping point where you need to have the domain knowledge to resolve the problem. even at the massive scales we are doing now, no AI can possibly hope to fit a repo into context and reason reliably about it, what we have are compromises and assumptions about where to focus the tool, slice up the input for it, and how to interpret its output. it's become yet another link in the fragile toolchain of complexity.
dunno how it works in other industries but adding "AI" to every project is not a panacea in the same way stuffing DEI bodies everywhere is not a panacea.

>>511991097
They're already zero for the big banks. The US needs more smaller banks with lower capital requirements.

nobody has shut the fuck up about AI ever since chatgpt came out in 2022 and we still haven't seen crazy productivity gains in any area.
even experienced software devs only see it as a moderate productivity booster.
despite billions of dollars and like gigawatts of electricity spent on it.
yeah I'm thinking the fed is right on this one.

>>511991374
Gas lighting was already very prevalent.

>>511991164
>Inductive reasoning is the least accurate anyway
I'd ask you to attempt to back that statement up, but I don't respect you enough to believe there's even a possibility that you could.

>>511985976 (OP)
>one-off invention like the lightbulb

Uh, you mean an invention so important and pivotal it is in every room of every building?

>>511991381
>average dipshit ChatGPT
You are talking about Apple, MIT and the Fed kek
b/c that is who wrote those papers

>>511991559
You posted Wikipedia with a huge banner about how it has major problems like lacking citations. You literally don't understand what you posted yet are strident that it was somehow proof of anything. Explain how LLMs actually work and then show how that will somehow lead to real reasoning, not just buffer space memory that it rescans over and over to try to reach a statistical accuracy rating high enough to shit it out to you.

>>511985976 (OP)
Obsolete systems often do try to save themselves. But tell me Anon. What's the first thing the lame man does when he doesn't need a crutch anymore?
https://youtu.be/tT0XXhfXYnU

>>511991414
>>511991436
AI hasn't been implemented yet and it won't be for a while
it's hard to say when will it become stable enough to do so, but my bayesian nutsack tingles when I predict 2030-2035

>>511991632
I'm agreeing with the papers chief. There is no path from LLM to AGI. The power usage is being subsidized, there's not going to be major efficiency gains other than ASIC type chips to lower energy usage but ultimately, it's a huge ML model with a giant DB that it has to run through x millions of times to produce an output that can be wrong.

>>511991684
>Hey guy, why don't you waste a bunch of time typing out all the things I'm too stupid to understand.

Yeah, ok I'll get right on that.

>>511992229
Yeah, you surely understand even though you fail basic logic that deduction is more accurate than induction, fucking retarded. Stop responding to me at this point, you're too stupid for this thread.

>>511991805
You're like one of those people who read about model collapse in a pop-sci article in 2023 and pretend it's destiny.

AlphaEvolve crushed any criticism Apple had. It literally made itself, and all software that runs on a GPU, more energy efficient.

>>511985976 (OP)
HOLY FUCK I LOVE AI NOW!!!
https://www.bitchute.com/video/74hmQ07CQMRx

>>511992431
Model collapse has been wrote about since 2011 at least, you're retarded. The collapse issue is very real and why they have to cater data for training or the model falls apart. They haven't found a way to fix that yet.
>AlphaEvolve
Fucking kill yourself, trying to pretend that model has anything to do with what we're discussing. Must be hard being that stupid.

>>511986467
I get all my piss in the toilet when I wake up in the middle of the night

>>511992422
>deduction is more accurate than induction
Deduct physics for me you Luddite.

>>511992715
>he doesn't use sonar
NGMI

>>511992229
lmao
you can take his posts, invert them, and see the future, tho
there's a clear path from LLMs to AGI
the most likely next stage is going to be small ML models that auto-train themselves until they reach beyond human-lvls of accuracy, which doesn't require a 300 IQ, just a 100-120 IQ that functions on a much smaller time/cost-scale than a human
(if you actually follow AI you can see this is the case)
(none of these retards follow AI at all)

>>511992422
lol that image

>>511991805
ok and thanks for the insight

>>511986271
>Reminds me of Krugman saying in 1998 that the internet was a nothingburger and would be less impactful than the fax machine.
Well the dot con bubble did pop two years later. Kek.

>>511992730
It's hilarious that you don't even understand basic logic, a branch of mathematics. Never change /pol/.

>>511992581
Show me that 2011 reference, big boy.
If you naively try to train an LLM on its own output, it will collapse, that's why you don't naively try to train an LLM on its own output. You can use something like DPO, or the multitude of other methods, that have been invented to train on synthetic data in the meantime.

>>511987626
>>511988015
never read any definition and certainly not anything they would publish, but it's what they're terming something that would resemble AI as the casual person has a perception of, something that isn't a chatbot entirely dependent on user input to generate output

woke right is just a term for miga types to be able to distance their zog agendas from the qucks in the republican party who are ironically waking up to what they're involved in
they didn't really think that one through

>>511993110
I'll have to hunt it down as I haven't seen it in years, let me know where your paper is in the meantime. Don't use Wikipedia, retard.

>>511986125
he's referring to somalis here

kek i'm really enjoying this

You guys really thought that they would let you use a misinformation tool forever especially when they want to make sure everyone's safe online? Silly billies, surely they didn't push AI on us to speed up this whole ID thing right? :)

<<<— the new pagan boogeyman
Alex Karp lost it (2/18/25) on CNBC and said "if you are not with tech and jews you were going to be left behind”
Listen to this audio of Karp on CNBC:
https://youtu.be/DkSs1y2uwVc?t=1796
Karp also called himself and his tech buddies the 'builders' of things in this country. when he talked, his stock dropped like a rock
Palantir sells ‘AI’ software to the CIA/FBI, etc.

>>511993677
Yeah yeah, googling takes a long time...
https://arxiv.org/pdf/2406.14532

>>511994058
>Training on model-generated synthetic data is a promising approach for finetuning LLMs, but it remains unclear when
it helps or hurts.
First sentence, you don't read shit.

>>511993997
Techno-Judeo-Christian alliance against the far right/left pagans is a hell of a narrative but actually makes more sense with the god complex they have

kek holy fuck

>>511994191
>Judeo-Christian
Judeo-Christian is an oxymoron

Judaism is the Old Testament (aka the Torah), it is about revenge, an eye for an eye.
Christianity is the New Testament (aka the Bible), it is about forgiveness, turn the other cheek.
Judaism is not the bearer of Christianity, but its antithesis.

>>511994316
I don’t make the rules, ZOG does

>>511988631
These seem really high, no?

>>511994369
this is just garden variety greed, nothing more

>>511993997
Alex Karp is a jew. Jews routinely use the "Judeo-Christian" bludgeon to convince Christians to love jews and Israel.

>>511988631
i don't believe a fucking WORD the Fed says

>>511994559
Out of 2.3 B Christians, there are ONLY 7M evangelicals that hope for jews to go back to israel, they follow the book of revelation, which the other 2.3 B Christians laugh at (pic related)

>>511986705
Now people can stay up gooning all night thank you lightbulb

>>511994127
Hey, maybe the rest of the paper contains some insight on how to train on synthetic data...

Oh, nevermind... the best part of the paper is always in the first sentence anyway, I don't even know what's the point of paragraphs is.

>>511994808
>evangelicals that hope for jews to go back to israel,
kek read the last paragraph
the jews said they can work with it if need be

>>511985976 (OP)
From the article
>they are finding scant evidence that many companies are actually using the technology.
>Data from job postings show only modest growth in demand for AI skills since 2017.
>questions remain about the quality and originality of AI-generated output.

>>511989976
We knew that last year.

https://www.sfgate.com/tech/article/analysts-google-meta-ai-hype-19563213.php
July 9, 2024
>In late June, analysts from Goldman Sachs, Barclays Investment Bank and venture giant Sequoia Capital each raised different versions of the same red flag. They argued in an interview, a note to clients and a blog post, respectively, that the mind-boggling amounts of money being dumped into AI will be very difficult to recoup. The technology’s development is extremely expensive, and its much-hyped benefits are totally unproven. In other words, some of the world’s largest companies and Silicon Valley’s best-known investors may be flushing cash down the drain.

https://www.goldmansachs.com/images/migrated/insights/pages/gs-research/gen-ai--too-much-spend%2C-too-little-benefit-/TOM_AI%202.0_ForRedaction.pdf

https://www.marketwatch.com/story/is-an-ai-stock-bubble-looming-thats-the-167-billion-question-53a76c00

https://www.sequoiacap.com/article/ais-600b-question/

>>511986792
It allows people to work in the dark nigger
>dark nigger

>>511986230
We still haven't recovered from the 2001 crash

>>511994848
You go ahead and break down your understanding of it, dumbass. They aren't drawing a conclusion in your cited paper. You're really fucking stupid. That paper only addresses mathematical accuracy of an LLM (fucking retarded), which is a huge problem with LLMs since they're language models anyway. Would be more sensible to just have a simple traffic steering proxy between the prompt and inference to steer from language to mathematical models.

Secondly, this is for fine tuning, which is not model tuning. They're taking a pretrained model and then fine tuning with synthetic data, I said training the model, not fine tuning the model. Tweaking a model that has already been trained is not the same as training an LLM model from scratch on synthetic data. Try again.

>>511991436
YEA BUT IT HELPS NIGGERS GRADUATE COLLEGE, THE TALENTED AFRICAN DOCTORS ARE SO THANKFUL!!! THE INDIANS ARE THANKING JEWS SAAR

Its a passing fad like the internet or the steam engine.

>>511991692
He breaks his leg again and petitions for extra time allotment
He laughs and turns the crutch into modern art
Genocides other 'lame men

>>511995800
kek
the ludditecost is gonna be a tonne of fun

>>511985976 (OP)
The banksters door want to lose their power to rule world to the technological IT giants.

>>511985976 (OP)
But the lightbulb made people more productive by being able to see better in dark places.

>>511997248
>to the technological IT giants.

afk

>>511992767
https://youtu.be/a05kgcI9D2Q